Thread by @JohnMullahy

Earlier threads have considered the use of the –recast– option in @Stata graphics. Here's another.

The –twoway function– command in Stata permits nice visualizations of explicit functions y=f(x) over some continuous domain of x-values. E.g.

twoway function y=normal(x), range(-3 3)

This can be helpful in...

— visualizing comparative features of different explicit functions

— visualizing theoretical vs. empirical results (e.g. goodness-of-fit)

— etc.

–twoway function– more or less assumes a continuous domain of x-values. The number of the function's evaluation points can be modified but this doesn't readily handle situations where the x-domain is naturally discrete.

This is where –recast– can help.

A general syntax for integer-valued x is (f(x) is some function Stata recognizes, eg normal(x)):

loc mn=(min x to plot)
loc mx=(max x to plot)
loc pts=`mx'-`mn'+1
tw fun y=f(x), ra(`mn' `mx') n(`pts') recast(scatter) [any options applicable to –twoway scatter– can go here]

For example, to see what a Poisson distribution with parameter lambda=2 looks like, one could use the command:

twoway function y=poissonp(2,x), ra(0 8) xlab(0(1)8) ylab(0(.1).3, ang(360)) xt(y) yt(Prob., orient(hor)) graphr(col(white))

which gives this:

Alternatively,

twoway function y=poissonp(2,x), ra(0 8) n(9) recast(scatter) msym(s) mcol(orange) msi(*2) xlab(0(1)8) ylab(0(.1).3, ang(360)) xt(y) yt(Prob., orient(hor)) graphr(col(white))

gives this:

As an alternative one could, of course, create a dataset that defines x at appropriate points, generate the corresponding y values as f(x), and then –scatter y x–. But why go to all that extra work when –recast– can do the heavy lifting?